Skip to contentSkip to conversation

Enable QEMU NVMe userspace driver on s390x

PATCH Superseded Farhan Ali 6 participants 18 messages Apr 17, 2025 – Apr 30, 2025

Subsystems (2)

v5 of 6 Superseded Latest is v6 →

lore
Share & tools
lore

Farhan Ali

author
#0 lore raw

v5 cover Enable QEMU NVMe userspace driver on s390x

Hi,

Recently on s390x we have enabled mmap support for vfio-pci devices [1]. This allows us to take advantage and use userspace drivers on s390x. However, on s390x we have special instructions for MMIO access. Starting with z15 (and newer platforms) we have new PCI Memory I/O (MIO) instructions which operate on virtually mapped PCI memory spaces, and can be used from userspace. On older platforms we would fallback to using existing system calls for MMIO access.

This patch series introduces support the PCI MIO instructions, and enables s390x support for the userspace NVMe driver on s390x. I would appreciate any review/feedback on the patches.

Thanks
Farhan

[1] https://lore.kernel.org/linux-s390/[email protected]/

ChangeLog
---------
v4 series https://lore.kernel.org/qemu-devel/[email protected]/
v4 -> v5
- Fixup typo in PCI MMIO API (patch 2).

v3 series https://lore.kernel.org/qemu-devel/[email protected]/
v3 -> v4
- Use generic ld/st functions for non s390x PCI access suggested by Alex (patch 2).
- Removed R-b for patch 2 as the host PCI MMIO access API changed for non-s390x. Would appreciate review on this again.

v2 series https://mail.gnu.org/archive/html/qemu-devel/2025-03/msg06847.html
v2 -> v3
- Update the PCI MMIO APIs to reflect that its PCI MMIO access on host
as suggested by Stefan(patch 2)
- Move s390x ifdef check to s390x_pci_mmio.h as suggested by Philippe (patch 1)
- Add R-bs for the respective patches.

v1 series https://mail.gnu.org/archive/html/qemu-devel/2025-03/msg06596.html
v1 -> v2
- Add 8 and 16 bit reads/writes for completeness (patch 1)
- Introduce new QEMU PCI MMIO read/write API as suggested by Stefan (patch 2)
- Update NVMe userspace driver to use QEMU PCI MMIO functions (patch 3)

Farhan Ali (3):
util: Add functions for s390x mmio read/write
include: Add a header to define host PCI MMIO functions block/nvme: Use host PCI MMIO API

 block/nvme.c                  |  41 +++++-----
 include/qemu/host-pci-mmio.h  | 141 ++++++++++++++++++++++++++++++++
 include/qemu/s390x_pci_mmio.h |  24 ++++++
 util/meson.build              |   2 +
 util/s390x_pci_mmio.c         | 148 ++++++++++++++++++++++++++++++++++
 5 files changed, 338 insertions(+), 18 deletions(-)
 create mode 100644 include/qemu/host-pci-mmio.h
 create mode 100644 include/qemu/s390x_pci_mmio.h
 create mode 100644 util/s390x_pci_mmio.c
Signature · 1 line
2.43.0
author
#1 lore raw

PATCH v5 1/3 util: Add functions for s390x mmio read/write

Starting with z15 (or newer) we can execute mmio
instructions from userspace. On older platforms
where we don't have these instructions available
we can fallback to using system calls to access
the PCI mapped resources.

This patch adds helper functions for mmio reads
and writes for s390x.

---
 include/qemu/s390x_pci_mmio.h |  24 ++++++
 util/meson.build              |   2 +
 util/s390x_pci_mmio.c         | 148 ++++++++++++++++++++++++++++++++++
 3 files changed, 174 insertions(+)
 create mode 100644 include/qemu/s390x_pci_mmio.h
 create mode 100644 util/s390x_pci_mmio.c
include/qemu/s390x_pci_mmio.h new file +24 -0
new file mode 100644
index 0000000000..c5f63ecefa
@@ -0,0 +1,24 @@
+/*
+ * s390x PCI MMIO definitions
+ *
+ * Copyright 2025 IBM Corp.
+ * Author(s): Farhan Ali <[email protected]>
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+#ifndef S390X_PCI_MMIO_H
+#define S390X_PCI_MMIO_H
+
+#ifdef __s390x__
+uint8_t s390x_pci_mmio_read_8(const void *ioaddr);
+uint16_t s390x_pci_mmio_read_16(const void *ioaddr);
+uint32_t s390x_pci_mmio_read_32(const void *ioaddr);
+uint64_t s390x_pci_mmio_read_64(const void *ioaddr);
+
+void s390x_pci_mmio_write_8(void *ioaddr, uint8_t val);
+void s390x_pci_mmio_write_16(void *ioaddr, uint16_t val);
+void s390x_pci_mmio_write_32(void *ioaddr, uint32_t val);
+void s390x_pci_mmio_write_64(void *ioaddr, uint64_t val);
+#endif /* __s390x__ */
+
+#endif /* S390X_PCI_MMIO_H */
util/meson.build +2 -0
index 780b5977a8..acb21592f9 100644
@@ -131,4 +131,6 @@ elif cpu in ['ppc', 'ppc64']
   util_ss.add(files('cpuinfo-ppc.c'))
 elif cpu in ['riscv32', 'riscv64']
   util_ss.add(files('cpuinfo-riscv.c'))
+elif cpu == 's390x'
+  util_ss.add(files('s390x_pci_mmio.c'))
 endif
util/s390x_pci_mmio.c new file +148 -0

148 lines not on this page · raw message

Signature · 1 line
2.43.0
#2 lore raw

Re: [PATCH v5 1/3] util: Add functions for s390x mmio read/write

Quoted 68 lines · Farhan Ali

Starting with z15 (or newer) we can execute mmio
instructions from userspace. On older platforms
where we don't have these instructions available
we can fallback to using system calls to access
the PCI mapped resources.

This patch adds helper functions for mmio reads
and writes for s390x.

---
  include/qemu/s390x_pci_mmio.h |  24 ++++++
  util/meson.build              |   2 +
  util/s390x_pci_mmio.c         | 148 ++++++++++++++++++++++++++++++++++
  3 files changed, 174 insertions(+)

create mode 100644 include/qemu/s390x_pci_mmio.h create mode 100644 util/s390x_pci_mmio.c

include/qemu/s390x_pci_mmio.h new file +24 -0
new file mode 100644
index 0000000000..c5f63ecefa
@@ -0,0 +1,24 @@
+/*
+ * s390x PCI MMIO definitions
+ *
+ * Copyright 2025 IBM Corp.
+ * Author(s): Farhan Ali <[email protected]>
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+#ifndef S390X_PCI_MMIO_H
+#define S390X_PCI_MMIO_H
+
+#ifdef __s390x__
+uint8_t s390x_pci_mmio_read_8(const void *ioaddr);
+uint16_t s390x_pci_mmio_read_16(const void *ioaddr);
+uint32_t s390x_pci_mmio_read_32(const void *ioaddr);
+uint64_t s390x_pci_mmio_read_64(const void *ioaddr);
+
+void s390x_pci_mmio_write_8(void *ioaddr, uint8_t val);
+void s390x_pci_mmio_write_16(void *ioaddr, uint16_t val);
+void s390x_pci_mmio_write_32(void *ioaddr, uint32_t val);
+void s390x_pci_mmio_write_64(void *ioaddr, uint64_t val);
+#endif /* __s390x__ */
+
+#endif /* S390X_PCI_MMIO_H */
util/meson.build +2 -0
index 780b5977a8..acb21592f9 100644
@@ -131,4 +131,6 @@ elif cpu in ['ppc', 'ppc64']
    util_ss.add(files('cpuinfo-ppc.c'))
  elif cpu in ['riscv32', 'riscv64']
    util_ss.add(files('cpuinfo-riscv.c'))
+elif cpu == 's390x'
+  util_ss.add(files('s390x_pci_mmio.c'))
  endif
util/s390x_pci_mmio.c new file +1 -0
new file mode 100644
index 0000000000..820458a026
@@ -0,0 +1,148 @@
+/*

util/s390x_pci_mmio.c line 11

Quoted 10 lines
util/s390x_pci_mmio.c new file +10 -0
+ * s390x PCI MMIO definitions
+ *
+ * Copyright 2025 IBM Corp.
+ * Author(s): Farhan Ali <[email protected]>
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#include "qemu/osdep.h"
+#include <unistd.h>

unistd.h is already included by osdep.h, so you don't have to include it again here.

Quoted 23 lines
util/s390x_pci_mmio.c new file +23 -0
+#include <sys/syscall.h>
+#include "qemu/s390x_pci_mmio.h"
+#include "elf.h"
+
+union register_pair {
+    unsigned __int128 pair;
+    struct {
+        uint64_t even;
+        uint64_t odd;
+    };
+};
+
+static bool is_mio_supported;
+
+static __attribute__((constructor)) void check_is_mio_supported(void)
+{
+    is_mio_supported = !!(qemu_getauxval(AT_HWCAP) & HWCAP_S390_PCI_MIO);
+}
+
+static uint64_t s390x_pcilgi(const void *ioaddr, size_t len)
+{
+    union register_pair ioaddr_len = { .even = (uint64_t)ioaddr,
+                                       .odd = len };

util/s390x_pci_mmio.c line 44

Quoted 10 lines
util/s390x_pci_mmio.c new file +10 -0
+    uint64_t val;
+    int cc;
+
+    asm volatile(
+        /* pcilgi */
+        ".insn   rre,0xb9d60000,%[val],%[ioaddr_len]\n"
+        "ipm     %[cc]\n"
+        "srl     %[cc],28\n"
+        : [cc] "=d"(cc), [val] "=d"(val),
+        [ioaddr_len] "+&d"(ioaddr_len.pair) :: "cc");

Do we need the "&" modifier here? ... at least the kernel does not seem to use it ...

Quoted 6 lines
util/s390x_pci_mmio.c new file +6 -0
+
+    if (cc) {
+        val = -1ULL;
+    }
+
+    return val;

util/s390x_pci_mmio.c line 60

Quoted 10 lines
util/s390x_pci_mmio.c new file +10 -0
+}
+
+static void s390x_pcistgi(void *ioaddr, uint64_t val, size_t len)
+{
+    union register_pair ioaddr_len = {.even = (uint64_t)ioaddr, .odd = len};
+
+    asm volatile (
+        /* pcistgi */
+        ".insn   rre,0xb9d40000,%[val],%[ioaddr_len]\n"
+        : [ioaddr_len] "+&d" (ioaddr_len.pair)

dito

Quoted 3 lines
util/s390x_pci_mmio.c new file +3 -0
+        : [val] "d" (val)
+        : "cc", "memory");
+}

...

util/s390x_pci_mmio.c line 148

Quoted 9 lines
util/s390x_pci_mmio.c new file +9 -0
+void s390x_pci_mmio_write_64(void *ioaddr, uint64_t val)
+{
+    if (is_mio_supported) {
+        s390x_pcistgi(ioaddr, val, sizeof(val));
+    } else {
+        syscall(__NR_s390_pci_mmio_write, ioaddr, &val, sizeof(val));
+    }
+}
+

FWIW, "git am" complains about "new blank line at EOF" here.

Apart from these nits, the patch looks sane to me.

Thomas

#3 lore raw

Re: [PATCH v5 1/3] util: Add functions for s390x mmio read/write

Quoted 111 lines · Thomas Huth
Quoted 68 lines · Farhan Ali

Starting with z15 (or newer) we can execute mmio
instructions from userspace. On older platforms
where we don't have these instructions available
we can fallback to using system calls to access
the PCI mapped resources.

This patch adds helper functions for mmio reads
and writes for s390x.

---
  include/qemu/s390x_pci_mmio.h |  24 ++++++
  util/meson.build              |   2 +
  util/s390x_pci_mmio.c         | 148 ++++++++++++++++++++++++++++++++++
  3 files changed, 174 insertions(+)

create mode 100644 include/qemu/s390x_pci_mmio.h create mode 100644 util/s390x_pci_mmio.c

include/qemu/s390x_pci_mmio.h new file +24 -0
new file mode 100644
index 0000000000..c5f63ecefa
@@ -0,0 +1,24 @@
+/*
+ * s390x PCI MMIO definitions
+ *
+ * Copyright 2025 IBM Corp.
+ * Author(s): Farhan Ali <[email protected]>
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+#ifndef S390X_PCI_MMIO_H
+#define S390X_PCI_MMIO_H
+
+#ifdef __s390x__
+uint8_t s390x_pci_mmio_read_8(const void *ioaddr);
+uint16_t s390x_pci_mmio_read_16(const void *ioaddr);
+uint32_t s390x_pci_mmio_read_32(const void *ioaddr);
+uint64_t s390x_pci_mmio_read_64(const void *ioaddr);
+
+void s390x_pci_mmio_write_8(void *ioaddr, uint8_t val);
+void s390x_pci_mmio_write_16(void *ioaddr, uint16_t val);
+void s390x_pci_mmio_write_32(void *ioaddr, uint32_t val);
+void s390x_pci_mmio_write_64(void *ioaddr, uint64_t val);
+#endif /* __s390x__ */
+
+#endif /* S390X_PCI_MMIO_H */
util/meson.build +2 -0
index 780b5977a8..acb21592f9 100644
@@ -131,4 +131,6 @@ elif cpu in ['ppc', 'ppc64']
    util_ss.add(files('cpuinfo-ppc.c'))
  elif cpu in ['riscv32', 'riscv64']
    util_ss.add(files('cpuinfo-riscv.c'))
+elif cpu == 's390x'
+  util_ss.add(files('s390x_pci_mmio.c'))
  endif
util/s390x_pci_mmio.c new file +1 -0
new file mode 100644
index 0000000000..820458a026
@@ -0,0 +1,148 @@
+/*

util/s390x_pci_mmio.c line 11

Quoted 10 lines
util/s390x_pci_mmio.c new file +10 -0
+ * s390x PCI MMIO definitions
+ *
+ * Copyright 2025 IBM Corp.
+ * Author(s): Farhan Ali <[email protected]>
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#include "qemu/osdep.h"
+#include <unistd.h>

unistd.h is already included by osdep.h, so you don't have to include it again here.

Quoted 28 lines
util/s390x_pci_mmio.c new file +28 -0
+#include <sys/syscall.h>
+#include "qemu/s390x_pci_mmio.h"
+#include "elf.h"
+
+union register_pair {
+    unsigned __int128 pair;
+    struct {
+        uint64_t even;
+        uint64_t odd;
+    };
+};
+
+static bool is_mio_supported;
+
+static __attribute__((constructor)) void check_is_mio_supported(void)
+{
+    is_mio_supported = !!(qemu_getauxval(AT_HWCAP) & HWCAP_S390_PCI_MIO);
+}
+
+static uint64_t s390x_pcilgi(const void *ioaddr, size_t len)
+{
+    union register_pair ioaddr_len = { .even = (uint64_t)ioaddr,
+                                       .odd = len };
+    uint64_t val;
+    int cc;
+
+    asm volatile(
+        /* pcilgi */
Quoted 9 lines

util/s390x_pci_mmio.c line 44

Quoted 5 lines
util/s390x_pci_mmio.c new file +5 -0
+        ".insn   rre,0xb9d60000,%[val],%[ioaddr_len]\n"
+        "ipm     %[cc]\n"
+        "srl     %[cc],28\n"
+        : [cc] "=d"(cc), [val] "=d"(val),
+        [ioaddr_len] "+&d"(ioaddr_len.pair) :: "cc");

Do we need the "&" modifier here? ... at least the kernel does not seem to use it ...

From my understanding it's not strictly needed, but I also used it in the rdma-core user-space code where I had pointed Farhan. I looked at the PR again but didn't find or remember why I added it. We do have the same constraint in the kernel implementation of the syscalll (__pcilg_mio_inuser()) that I worked on in the same time frame but that also has other instructions before the PCILGI. I see it in other uses of "union register_pair" in the kernel too. Reading the gcc docs again, I also think it does give gcc the right information here. The address/length pair is only written to after it is read what it should/can do with this information I don't know. Adding Heiko who knows way more than I about inline assembly.

Quoted 22 lines
Quoted 6 lines
util/s390x_pci_mmio.c new file +6 -0
+
+    if (cc) {
+        val = -1ULL;
+    }
+
+    return val;

util/s390x_pci_mmio.c line 60

Quoted 10 lines
util/s390x_pci_mmio.c new file +10 -0
+}
+
+static void s390x_pcistgi(void *ioaddr, uint64_t val, size_t len)
+{
+    union register_pair ioaddr_len = {.even = (uint64_t)ioaddr, .odd = len};
+
+    asm volatile (
+        /* pcistgi */
+        ".insn   rre,0xb9d40000,%[val],%[ioaddr_len]\n"
+        : [ioaddr_len] "+&d" (ioaddr_len.pair)

dito

Quoted 3 lines
util/s390x_pci_mmio.c new file +3 -0
+        : [val] "d" (val)
+        : "cc", "memory");
+}
#4 lore raw

Re: [PATCH v5 1/3] util: Add functions for s390x mmio read/write

Quoted 10 lines · Niklas Schnelle
Quoted 9 lines · Thomas Huth
Quoted 7 lines · Farhan Ali
util/s390x_pci_mmio.c new file +7 -0
+    asm volatile(
+        /* pcilgi */
+        ".insn   rre,0xb9d60000,%[val],%[ioaddr_len]\n"
+        "ipm     %[cc]\n"
+        "srl     %[cc],28\n"
+        : [cc] "=d"(cc), [val] "=d"(val),
+        [ioaddr_len] "+&d"(ioaddr_len.pair) :: "cc");
Quoted 5 lines
Quoted 2 lines

Do we need the "&" modifier here? ... at least the kernel does not seem to use it ...

From my understanding it's not strictly needed, but I also used it in the rdma-core user-space code where I had pointed Farhan. I looked at

It is not needed, since all inputs are consumed before to any output is written to.

Quoted 6 lines
Quoted 6 lines

util/s390x_pci_mmio.c line 60

Quoted 4 lines
util/s390x_pci_mmio.c new file +4 -0
+    asm volatile (
+        /* pcistgi */
+        ".insn   rre,0xb9d40000,%[val],%[ioaddr_len]\n"
+        : [ioaddr_len] "+&d" (ioaddr_len.pair)

dito

Same here, it is not needed.

author
#5 lore raw

Re: [PATCH v5 1/3] util: Add functions for s390x mmio read/write

Quoted 10 lines · Heiko Carstens
Quoted 9 lines · Niklas Schnelle
Quoted 8 lines · Thomas Huth
Quoted 7 lines · Farhan Ali
util/s390x_pci_mmio.c new file +7 -0
+    asm volatile(
+        /* pcilgi */
+        ".insn   rre,0xb9d60000,%[val],%[ioaddr_len]\n"
+        "ipm     %[cc]\n"
+        "srl     %[cc],28\n"
+        : [cc] "=d"(cc), [val] "=d"(val),
+        [ioaddr_len] "+&d"(ioaddr_len.pair) :: "cc");
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

Do we need the "&" modifier here? ... at least the kernel does not seem to use it ...

From my understanding it's not strictly needed, but I also used it in
the rdma-core user-space code where I had pointed Farhan. I looked at

It is not needed, since all inputs are consumed before to any output is written to.

Ack, will update the patch.

Quoted 6 lines
Quoted 5 lines
Quoted 5 lines

util/s390x_pci_mmio.c line 60

Quoted 4 lines
util/s390x_pci_mmio.c new file +4 -0
+    asm volatile (
+        /* pcistgi */
+        ".insn   rre,0xb9d40000,%[val],%[ioaddr_len]\n"
+        : [ioaddr_len] "+&d" (ioaddr_len.pair)

dito

Same here, it is not needed.

Ack, will update the patch here as well.

Thanks
Farhan

author
#6 lore raw

PATCH v5 2/3 include: Add a header to define host PCI MMIO functions

In reply to Farhan Ali

Add a generic API for host PCI MMIO reads/writes (e.g. Linux VFIO BAR accesses). The functions access little endian memory and returns the result in host cpu endianness.

---
 include/qemu/host-pci-mmio.h | 141 +++++++++++++++++++++++++++++++++++
 1 file changed, 141 insertions(+)
 create mode 100644 include/qemu/host-pci-mmio.h
include/qemu/host-pci-mmio.h new file +141 -0
new file mode 100644
index 0000000000..c93f77dcd4
@@ -0,0 +1,141 @@
+/*
+ * API for host PCI MMIO accesses (e.g. Linux VFIO BARs)
+ *
+ * Copyright 2025 IBM Corp.
+ * Author(s): Farhan Ali <[email protected]>
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#ifndef HOST_PCI_MMIO_H
+#define HOST_PCI_MMIO_H
+
+#include "qemu/bswap.h"
+#include "qemu/s390x_pci_mmio.h"
+
+
+static inline uint8_t host_pci_ldub_p(const void *ioaddr)
+{
+    uint8_t ret = 0;
+#ifdef __s390x__
+    ret = s390x_pci_mmio_read_8(ioaddr);
+#else
+    ret = ldub_p(ioaddr);
+#endif
+
+    return ret;
+}
+
+static inline uint16_t host_pci_lduw_le_p(const void *ioaddr)
+{
+    uint16_t ret = 0;
+#ifdef __s390x__
+    ret = le16_to_cpu(s390x_pci_mmio_read_16(ioaddr));
+#else
+    ret = lduw_le_p(ioaddr);
+#endif
+
+    return ret;
+}
+
+static inline uint32_t host_pci_ldl_le_p(const void *ioaddr)
+{
+    uint32_t ret = 0;
+#ifdef __s390x__
+    ret = le32_to_cpu(s390x_pci_mmio_read_32(ioaddr));
+#else
+    ret = (uint32_t)ldl_le_p(ioaddr);
+#endif
+
+    return ret;
+}
+
+static inline uint64_t host_pci_ldq_le_p(const void *ioaddr)
+{
+    uint64_t ret = 0;
+#ifdef __s390x__
+    ret = le64_to_cpu(s390x_pci_mmio_read_64(ioaddr));
+#else
+    ret = ldq_le_p(ioaddr);
+#endif
+
+    return ret;
+}
+
+static inline void host_pci_stb_le_p(void *ioaddr, uint8_t val)
+{
+
+#ifdef __s390x__
+    s390x_pci_mmio_write_8(ioaddr, val);
+#else
+    stb_p(ioaddr, val);
+#endif
+}
+
+static inline void host_pci_stw_le_p(void *ioaddr, uint16_t val)
+{
+
+#ifdef __s390x__
+    s390x_pci_mmio_write_16(ioaddr, cpu_to_le16(val));
+#else
+    stw_le_p(ioaddr, val);
+#endif
+}
+
+static inline void host_pci_stl_le_p(void *ioaddr, uint32_t val)
+{
+
+#ifdef __s390x__
+    s390x_pci_mmio_write_32(ioaddr, cpu_to_le32(val));
+#else
+    stl_le_p(ioaddr, val);
+#endif
+}
+
+static inline void host_pci_stq_le_p(void *ioaddr, uint64_t val)
+{
+
+#ifdef __s390x__
+    s390x_pci_mmio_write_64(ioaddr, cpu_to_le64(val));
+#else
+    stq_le_p(ioaddr, val);
+#endif
+}
+
+static inline uint64_t host_pci_ldn_le_p(const void *ioaddr, int sz)
+{
+    switch (sz) {
+    case 1:
+        return host_pci_ldub_p(ioaddr);
+    case 2:
+        return host_pci_lduw_le_p(ioaddr);
+    case 4:
+        return host_pci_ldl_le_p(ioaddr);
+    case 8:
+        return host_pci_ldq_le_p(ioaddr);
+    default:
+        g_assert_not_reached();
+    }
+}
+
+static inline void host_pci_stn_le_p(void *ioaddr, int sz, uint64_t v)
+{
+    switch (sz) {
+    case 1:
+        host_pci_stb_le_p(ioaddr, v);
+        break;
+    case 2:
+        host_pci_stw_le_p(ioaddr, v);
+        break;
+    case 4:
+        host_pci_stl_le_p(ioaddr, v);
+        break;
+    case 8:
+        host_pci_stq_le_p(ioaddr, v);
+        break;
+    default:
+        g_assert_not_reached();
+    }
+}
+
+#endif
Signature · 1 line
2.43.0
Reviewed-by on v5 2/3
#7 lore raw

Re: [PATCH v5 2/3] include: Add a header to define host PCI MMIO functions

Quoted 10 lines · Farhan Ali

Add a generic API for host PCI MMIO reads/writes (e.g. Linux VFIO BAR accesses). The functions access little endian memory and returns the result in host cpu endianness.

---
 include/qemu/host-pci-mmio.h | 141 +++++++++++++++++++++++++++++++++++
 1 file changed, 141 insertions(+)
 create mode 100644 include/qemu/host-pci-mmio.h
#8 lore raw

Re: [PATCH v5 2/3] include: Add a header to define host PCI MMIO functions

In reply to Farhan Ali
Quoted 23 lines · Farhan Ali

Add a generic API for host PCI MMIO reads/writes (e.g. Linux VFIO BAR accesses). The functions access little endian memory and returns the result in host cpu endianness.

---
  include/qemu/host-pci-mmio.h | 141 +++++++++++++++++++++++++++++++++++
  1 file changed, 141 insertions(+)

create mode 100644 include/qemu/host-pci-mmio.h

include/qemu/host-pci-mmio.h new file +6 -0
new file mode 100644
index 0000000000..c93f77dcd4
@@ -0,0 +1,141 @@
+/*
+ * API for host PCI MMIO accesses (e.g. Linux VFIO BARs)
+ *
+ * Copyright 2025 IBM Corp.
+ * Author(s): Farhan Ali <[email protected]>
+ *

include/qemu/host-pci-mmio.h line 16

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#ifndef HOST_PCI_MMIO_H
+#define HOST_PCI_MMIO_H
+
+#include "qemu/bswap.h"
+#include "qemu/s390x_pci_mmio.h"
+
+

Cosmetic nit: It's more common in QEMU to only use one empty line instead of two.

Quoted 21 lines
include/qemu/host-pci-mmio.h new file +21 -0
+static inline uint8_t host_pci_ldub_p(const void *ioaddr)
+{
+    uint8_t ret = 0;
+#ifdef __s390x__
+    ret = s390x_pci_mmio_read_8(ioaddr);
+#else
+    ret = ldub_p(ioaddr);
+#endif
+
+    return ret;
+}
+
+static inline uint16_t host_pci_lduw_le_p(const void *ioaddr)
+{
+    uint16_t ret = 0;
+#ifdef __s390x__
+    ret = le16_to_cpu(s390x_pci_mmio_read_16(ioaddr));
+#else
+    ret = lduw_le_p(ioaddr);
+#endif
+

include/qemu/host-pci-mmio.h line 47

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+    return ret;
+}
+
+static inline uint32_t host_pci_ldl_le_p(const void *ioaddr)
+{
+    uint32_t ret = 0;
+#ifdef __s390x__
+    ret = le32_to_cpu(s390x_pci_mmio_read_32(ioaddr));
+#else
+    ret = (uint32_t)ldl_le_p(ioaddr);

This is the only spot where you used a cast. Is it necessary, or could it be omitted?

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+#endif
+
+    return ret;
+}
+
+static inline uint64_t host_pci_ldq_le_p(const void *ioaddr)
+{
+    uint64_t ret = 0;
+#ifdef __s390x__
+    ret = le64_to_cpu(s390x_pci_mmio_read_64(ioaddr));

include/qemu/host-pci-mmio.h line 67

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+#else
+    ret = ldq_le_p(ioaddr);
+#endif
+
+    return ret;
+}
+
+static inline void host_pci_stb_le_p(void *ioaddr, uint8_t val)
+{
+

Remove the empty line, please.

include/qemu/host-pci-mmio.h line 77

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+#ifdef __s390x__
+    s390x_pci_mmio_write_8(ioaddr, val);
+#else
+    stb_p(ioaddr, val);
+#endif
+}
+
+static inline void host_pci_stw_le_p(void *ioaddr, uint16_t val)
+{
+

dito.

include/qemu/host-pci-mmio.h line 87

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+#ifdef __s390x__
+    s390x_pci_mmio_write_16(ioaddr, cpu_to_le16(val));
+#else
+    stw_le_p(ioaddr, val);
+#endif
+}
+
+static inline void host_pci_stl_le_p(void *ioaddr, uint32_t val)
+{
+

dito.

include/qemu/host-pci-mmio.h line 97

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+#ifdef __s390x__
+    s390x_pci_mmio_write_32(ioaddr, cpu_to_le32(val));
+#else
+    stl_le_p(ioaddr, val);
+#endif
+}
+
+static inline void host_pci_stq_le_p(void *ioaddr, uint64_t val)
+{
+

dito

Quoted 34 lines
include/qemu/host-pci-mmio.h new file +34 -0
+#ifdef __s390x__
+    s390x_pci_mmio_write_64(ioaddr, cpu_to_le64(val));
+#else
+    stq_le_p(ioaddr, val);
+#endif
+}
+
+static inline uint64_t host_pci_ldn_le_p(const void *ioaddr, int sz)
+{
+    switch (sz) {
+    case 1:
+        return host_pci_ldub_p(ioaddr);
+    case 2:
+        return host_pci_lduw_le_p(ioaddr);
+    case 4:
+        return host_pci_ldl_le_p(ioaddr);
+    case 8:
+        return host_pci_ldq_le_p(ioaddr);
+    default:
+        g_assert_not_reached();
+    }
+}
+
+static inline void host_pci_stn_le_p(void *ioaddr, int sz, uint64_t v)
+{
+    switch (sz) {
+    case 1:
+        host_pci_stb_le_p(ioaddr, v);
+        break;
+    case 2:
+        host_pci_stw_le_p(ioaddr, v);
+        break;
+    case 4:
+        host_pci_stl_le_p(ioaddr, v);

include/qemu/host-pci-mmio.h line 141

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+        break;
+    case 8:
+        host_pci_stq_le_p(ioaddr, v);
+        break;
+    default:
+        g_assert_not_reached();
+    }
+}
+
+#endif

Apart from the nits, patch looks good to me.

Thomas

author
#9 lore raw

Re: [PATCH v5 2/3] include: Add a header to define host PCI MMIO functions

..snip...

Quoted 9 lines

include/qemu/host-pci-mmio.h line 47

Quoted 7 lines
include/qemu/host-pci-mmio.h new file +7 -0
+static inline uint32_t host_pci_ldl_le_p(const void *ioaddr)
+{
+    uint32_t ret = 0;
+#ifdef __s390x__
+    ret = le32_to_cpu(s390x_pci_mmio_read_32(ioaddr));
+#else
+    ret = (uint32_t)ldl_le_p(ioaddr);

This is the only spot where you used a cast. Is it necessary, or could it be omitted?

Yes, the ldl_le_p returns an int. We do similar cast here https://github.com/qemu/qemu/blob/73d29ea2417b58ca55fba1aa468ba38e3607b583/include/qemu/bswap.h#L416

Quoted 110 lines
Quoted 11 lines
include/qemu/host-pci-mmio.h new file +10 -0
+#endif
+
+    return ret;
+}
+
+static inline uint64_t host_pci_ldq_le_p(const void *ioaddr)
+{
+    uint64_t ret = 0;
+#ifdef __s390x__
+    ret = le64_to_cpu(s390x_pci_mmio_read_64(ioaddr));

include/qemu/host-pci-mmio.h line 67

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+#else
+    ret = ldq_le_p(ioaddr);
+#endif
+
+    return ret;
+}
+
+static inline void host_pci_stb_le_p(void *ioaddr, uint8_t val)
+{
+

Remove the empty line, please.

include/qemu/host-pci-mmio.h line 77

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+#ifdef __s390x__
+    s390x_pci_mmio_write_8(ioaddr, val);
+#else
+    stb_p(ioaddr, val);
+#endif
+}
+
+static inline void host_pci_stw_le_p(void *ioaddr, uint16_t val)
+{
+

dito.

include/qemu/host-pci-mmio.h line 87

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+#ifdef __s390x__
+    s390x_pci_mmio_write_16(ioaddr, cpu_to_le16(val));
+#else
+    stw_le_p(ioaddr, val);
+#endif
+}
+
+static inline void host_pci_stl_le_p(void *ioaddr, uint32_t val)
+{
+

dito.

include/qemu/host-pci-mmio.h line 97

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+#ifdef __s390x__
+    s390x_pci_mmio_write_32(ioaddr, cpu_to_le32(val));
+#else
+    stl_le_p(ioaddr, val);
+#endif
+}
+
+static inline void host_pci_stq_le_p(void *ioaddr, uint64_t val)
+{
+

dito

Quoted 34 lines
include/qemu/host-pci-mmio.h new file +34 -0
+#ifdef __s390x__
+    s390x_pci_mmio_write_64(ioaddr, cpu_to_le64(val));
+#else
+    stq_le_p(ioaddr, val);
+#endif
+}
+
+static inline uint64_t host_pci_ldn_le_p(const void *ioaddr, int sz)
+{
+    switch (sz) {
+    case 1:
+        return host_pci_ldub_p(ioaddr);
+    case 2:
+        return host_pci_lduw_le_p(ioaddr);
+    case 4:
+        return host_pci_ldl_le_p(ioaddr);
+    case 8:
+        return host_pci_ldq_le_p(ioaddr);
+    default:
+        g_assert_not_reached();
+    }
+}
+
+static inline void host_pci_stn_le_p(void *ioaddr, int sz, uint64_t v)
+{
+    switch (sz) {
+    case 1:
+        host_pci_stb_le_p(ioaddr, v);
+        break;
+    case 2:
+        host_pci_stw_le_p(ioaddr, v);
+        break;
+    case 4:
+        host_pci_stl_le_p(ioaddr, v);

include/qemu/host-pci-mmio.h line 141

Quoted 10 lines
include/qemu/host-pci-mmio.h new file +10 -0
+        break;
+    case 8:
+        host_pci_stq_le_p(ioaddr, v);
+        break;
+    default:
+        g_assert_not_reached();
+    }
+}
+
+#endif

Apart from the nits, patch looks good to me.

Quoted 1 line

 Thomas

Thanks for reviewing! will fix the nits in the next revision.

Thanks

Farhan

#10 lore raw

Re: [PATCH v5 2/3] include: Add a header to define host PCI MMIO functions

Quoted 13 lines · Farhan Ali

..snip...

Quoted 9 lines

include/qemu/host-pci-mmio.h line 47

Quoted 7 lines
include/qemu/host-pci-mmio.h new file +7 -0
+static inline uint32_t host_pci_ldl_le_p(const void *ioaddr)
+{
+    uint32_t ret = 0;
+#ifdef __s390x__
+    ret = le32_to_cpu(s390x_pci_mmio_read_32(ioaddr));
+#else
+    ret = (uint32_t)ldl_le_p(ioaddr);

This is the only spot where you used a cast. Is it necessary, or could it be omitted?

Quoted 1 line

Yes, the ldl_le_p returns an int. We do similar cast here https:// github.com/qemu/qemu/blob/73d29ea2417b58ca55fba1aa468ba38e3607b583/include/ qemu/bswap.h#L416

... but that function there returns an 64-bit value, while you are assigning the value to a 32-bit variable here, and you also only return a 32-bit value from the function here. So there is no way that this could accidentally be sign-extended, could it?

Thomas

author
#11 lore raw

Re: [PATCH v5 2/3] include: Add a header to define host PCI MMIO functions

Quoted 14 lines · Thomas Huth
Quoted 13 lines · Farhan Ali

..snip...

Quoted 9 lines

include/qemu/host-pci-mmio.h line 47

Quoted 7 lines
include/qemu/host-pci-mmio.h new file +7 -0
+static inline uint32_t host_pci_ldl_le_p(const void *ioaddr)
+{
+    uint32_t ret = 0;
+#ifdef __s390x__
+    ret = le32_to_cpu(s390x_pci_mmio_read_32(ioaddr));
+#else
+    ret = (uint32_t)ldl_le_p(ioaddr);

This is the only spot where you used a cast. Is it necessary, or could it be omitted?

Quoted 7 lines
Quoted 1 line

Yes, the ldl_le_p returns an int. We do similar cast here https:// github.com/qemu/qemu/blob/73d29ea2417b58ca55fba1aa468ba38e3607b583/include/ qemu/bswap.h#L416

... but that function there returns an 64-bit value, while you are assigning the value to a 32-bit variable here, and you also only return a 32-bit value from the function here. So there is no way that this could accidentally be sign-extended, could it?

 Thomas

I checked the Linux kernel and there too we seem to be returning unsigned 32 bit values (readl) for 32 bit mmio reads. So I think it should be safe. Do you think we should remove the cast?

Thanks

Farhan

#12 lore raw

Re: [PATCH v5 2/3] include: Add a header to define host PCI MMIO functions

Quoted 16 lines · Farhan Ali
Quoted 14 lines · Thomas Huth
Quoted 13 lines · Farhan Ali

..snip...

Quoted 9 lines

include/qemu/host-pci-mmio.h line 47

Quoted 7 lines
include/qemu/host-pci-mmio.h new file +7 -0
+static inline uint32_t host_pci_ldl_le_p(const void *ioaddr)
+{
+    uint32_t ret = 0;
+#ifdef __s390x__
+    ret = le32_to_cpu(s390x_pci_mmio_read_32(ioaddr));
+#else
+    ret = (uint32_t)ldl_le_p(ioaddr);

This is the only spot where you used a cast. Is it necessary, or could it be omitted?

Quoted 8 lines
Quoted 7 lines
Quoted 1 line

Yes, the ldl_le_p returns an int. We do similar cast here https:// github.com/qemu/qemu/blob/73d29ea2417b58ca55fba1aa468ba38e3607b583/ include/ qemu/bswap.h#L416

... but that function there returns an 64-bit value, while you are assigning the value to a 32-bit variable here, and you also only return a 32-bit value from the function here. So there is no way that this could accidentally be sign-extended, could it?

 Thomas

I checked the Linux kernel and there too we seem to be returning unsigned 32 bit values (readl) for 32 bit mmio reads. So I think it should be safe. Do you think we should remove the cast?

It looks redundant to me, so I'd vote to remove it, yes!

Thomas

author
#13 lore raw

PATCH v5 3/3 block/nvme: Use host PCI MMIO API

In reply to Farhan Ali

Use the host PCI MMIO functions to read/write to NVMe registers, rather than directly accessing them.

---
 block/nvme.c | 41 +++++++++++++++++++++++------------------
 1 file changed, 23 insertions(+), 18 deletions(-)
block/nvme.c +23 -18
index bbf7c23dcd..8df53ee4ca 100644
@@ -18,6 +18,7 @@
 #include "qobject/qstring.h"
 #include "qemu/defer-call.h"
 #include "qemu/error-report.h"
+#include "qemu/host-pci-mmio.h"
 #include "qemu/main-loop.h"
 #include "qemu/module.h"
 #include "qemu/cutils.h"
@@ -60,7 +61,7 @@ typedef struct {
     uint8_t  *queue;
     uint64_t iova;
     /* Hardware MMIO register */
-    volatile uint32_t *doorbell;
+    uint32_t *doorbell;
 } NVMeQueue;
 
 typedef struct {
@@ -100,7 +101,7 @@ struct BDRVNVMeState {
     QEMUVFIOState *vfio;
     void *bar0_wo_map;
     /* Memory mapped registers */
-    volatile struct {
+    struct {
         uint32_t sq_tail;
         uint32_t cq_head;
     } *doorbells;
@@ -292,7 +293,7 @@ static void nvme_kick(NVMeQueuePair *q)
     assert(!(q->sq.tail & 0xFF00));
     /* Fence the write to submission queue entry before notifying the device. */
     smp_wmb();
-    *q->sq.doorbell = cpu_to_le32(q->sq.tail);
+    host_pci_stl_le_p(q->sq.doorbell, q->sq.tail);
     q->inflight += q->need_kick;
     q->need_kick = 0;
 }
@@ -441,7 +442,7 @@ static bool nvme_process_completion(NVMeQueuePair *q)
     if (progress) {
         /* Notify the device so it can post more completions. */
         smp_mb_release();
-        *q->cq.doorbell = cpu_to_le32(q->cq.head);
+        host_pci_stl_le_p(q->cq.doorbell, q->cq.head);
         nvme_wake_free_req_locked(q);
     }
 
@@ -460,7 +461,7 @@ static void nvme_process_completion_bh(void *opaque)
      * so notify the device that it has space to fill in more completions now.
      */
     smp_mb_release();
-    *q->cq.doorbell = cpu_to_le32(q->cq.head);
+    host_pci_stl_le_p(q->cq.doorbell, q->cq.head);
     nvme_wake_free_req_locked(q);
 
     nvme_process_completion(q);
@@ -749,9 +750,10 @@ static int nvme_init(BlockDriverState *bs, const char *device, int namespace,
     int ret;
     uint64_t cap;
     uint32_t ver;
+    uint32_t cc;
     uint64_t timeout_ms;
     uint64_t deadline, now;
-    volatile NvmeBar *regs = NULL;
+    NvmeBar *regs = NULL;
 
     qemu_co_mutex_init(&s->dma_map_lock);
     qemu_co_queue_init(&s->dma_flush_queue);
@@ -779,7 +781,7 @@ static int nvme_init(BlockDriverState *bs, const char *device, int namespace,
     /* Perform initialize sequence as described in NVMe spec "7.6.1
      * Initialization". */
 
-    cap = le64_to_cpu(regs->cap);
+    cap = host_pci_ldq_le_p(&regs->cap);
     trace_nvme_controller_capability_raw(cap);
     trace_nvme_controller_capability("Maximum Queue Entries Supported",
                                      1 + NVME_CAP_MQES(cap));
@@ -805,16 +807,17 @@ static int nvme_init(BlockDriverState *bs, const char *device, int namespace,
     bs->bl.request_alignment = s->page_size;
     timeout_ms = MIN(500 * NVME_CAP_TO(cap), 30000);
 
-    ver = le32_to_cpu(regs->vs);
+    ver = host_pci_ldl_le_p(&regs->vs);
     trace_nvme_controller_spec_version(extract32(ver, 16, 16),
                                        extract32(ver, 8, 8),
                                        extract32(ver, 0, 8));
 
     /* Reset device to get a clean state. */
-    regs->cc = cpu_to_le32(le32_to_cpu(regs->cc) & 0xFE);
+    cc = host_pci_ldl_le_p(&regs->cc);
+    host_pci_stl_le_p(&regs->cc, cc & 0xFE);
     /* Wait for CSTS.RDY = 0. */
     deadline = qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + timeout_ms * SCALE_MS;
-    while (NVME_CSTS_RDY(le32_to_cpu(regs->csts))) {
+    while (NVME_CSTS_RDY(host_pci_ldl_le_p(&regs->csts))) {
         if (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) > deadline) {
             error_setg(errp, "Timeout while waiting for device to reset (%"
                              PRId64 " ms)",
@@ -843,19 +846,21 @@ static int nvme_init(BlockDriverState *bs, const char *device, int namespace,
     s->queues[INDEX_ADMIN] = q;
     s->queue_count = 1;
     QEMU_BUILD_BUG_ON((NVME_QUEUE_SIZE - 1) & 0xF000);
-    regs->aqa = cpu_to_le32(((NVME_QUEUE_SIZE - 1) << AQA_ACQS_SHIFT) |
-                            ((NVME_QUEUE_SIZE - 1) << AQA_ASQS_SHIFT));
-    regs->asq = cpu_to_le64(q->sq.iova);
-    regs->acq = cpu_to_le64(q->cq.iova);
+    host_pci_stl_le_p(&regs->aqa,
+                        ((NVME_QUEUE_SIZE - 1) << AQA_ACQS_SHIFT) |
+                        ((NVME_QUEUE_SIZE - 1) << AQA_ASQS_SHIFT));
+    host_pci_stq_le_p(&regs->asq, q->sq.iova);
+    host_pci_stq_le_p(&regs->acq, q->cq.iova);
 
     /* After setting up all control registers we can enable device now. */
-    regs->cc = cpu_to_le32((ctz32(NVME_CQ_ENTRY_BYTES) << CC_IOCQES_SHIFT) |
-                           (ctz32(NVME_SQ_ENTRY_BYTES) << CC_IOSQES_SHIFT) |
-                           CC_EN_MASK);
+    host_pci_stl_le_p(&regs->cc,
+                      (ctz32(NVME_CQ_ENTRY_BYTES) << CC_IOCQES_SHIFT) |
+                      (ctz32(NVME_SQ_ENTRY_BYTES) << CC_IOSQES_SHIFT) |
+                      CC_EN_MASK);
     /* Wait for CSTS.RDY = 1. */
     now = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
     deadline = now + timeout_ms * SCALE_MS;
-    while (!NVME_CSTS_RDY(le32_to_cpu(regs->csts))) {
+    while (!NVME_CSTS_RDY(host_pci_ldl_le_p(&regs->csts))) {
         if (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) > deadline) {
             error_setg(errp, "Timeout while waiting for device to start (%"
                              PRId64 " ms)",
Signature · 1 line
2.43.0
Reviewed-by on v5 3/3
#14 lore raw

Re: [PATCH v5 3/3] block/nvme: Use host PCI MMIO API

Quoted 10 lines · Farhan Ali

Use the host PCI MMIO functions to read/write to NVMe registers, rather than directly accessing them.

---
  block/nvme.c | 41 +++++++++++++++++++++++------------------
  1 file changed, 23 insertions(+), 18 deletions(-)
author
#15 lore raw
In reply to Farhan Ali

Hi Alex,

Polite ping. Please let me know if there are any concerns with this version of the patches.

Thanks

Farhan

Quoted 59 lines · Farhan Ali

Hi,

Recently on s390x we have enabled mmap support for vfio-pci devices [1]. This allows us to take advantage and use userspace drivers on s390x. However, on s390x we have special instructions for MMIO access. Starting with z15 (and newer platforms) we have new PCI Memory I/O (MIO) instructions which operate on virtually mapped PCI memory spaces, and can be used from userspace. On older platforms we would fallback to using existing system calls for MMIO access.

This patch series introduces support the PCI MIO instructions, and enables s390x support for the userspace NVMe driver on s390x. I would appreciate any review/feedback on the patches.

Thanks
Farhan

[1] https://lore.kernel.org/linux-s390/[email protected]/

ChangeLog
---------
v4 series https://lore.kernel.org/qemu-devel/[email protected]/
v4 -> v5
- Fixup typo in PCI MMIO API (patch 2).

v3 series https://lore.kernel.org/qemu-devel/[email protected]/
v3 -> v4
- Use generic ld/st functions for non s390x PCI access suggested by Alex (patch 2).
- Removed R-b for patch 2 as the host PCI MMIO access API changed for non-s390x. Would appreciate review on this again.

v2 series https://mail.gnu.org/archive/html/qemu-devel/2025-03/msg06847.html
v2 -> v3
- Update the PCI MMIO APIs to reflect that its PCI MMIO access on host
as suggested by Stefan(patch 2)
- Move s390x ifdef check to s390x_pci_mmio.h as suggested by Philippe (patch 1)
- Add R-bs for the respective patches.

v1 series https://mail.gnu.org/archive/html/qemu-devel/2025-03/msg06596.html
v1 -> v2
- Add 8 and 16 bit reads/writes for completeness (patch 1)
- Introduce new QEMU PCI MMIO read/write API as suggested by Stefan (patch 2)
- Update NVMe userspace driver to use QEMU PCI MMIO functions (patch 3)

Farhan Ali (3):
util: Add functions for s390x mmio read/write
include: Add a header to define host PCI MMIO functions block/nvme: Use host PCI MMIO API

  block/nvme.c                  |  41 +++++-----
  include/qemu/host-pci-mmio.h  | 141 ++++++++++++++++++++++++++++++++
  include/qemu/s390x_pci_mmio.h |  24 ++++++
  util/meson.build              |   2 +
  util/s390x_pci_mmio.c         | 148 ++++++++++++++++++++++++++++++++++
  5 files changed, 338 insertions(+), 18 deletions(-)

create mode 100644 include/qemu/host-pci-mmio.h create mode 100644 include/qemu/s390x_pci_mmio.h create mode 100644 util/s390x_pci_mmio.c

#16 lore raw
Quoted 3 lines · Farhan Ali

Hi Alex,

Polite ping. Please let me know if there are any concerns with this version of the patches.

I assumed that Stefan would pick the patches up for the nvme/block tree, but I can also take them through the s390x tree instead if that's preferred. But please address/comment on the nits that I found first.

Thanks,
Thomas

#17 lore raw
In reply to Farhan Ali
Quoted 1 line · Farhan Ali

Hi Alex,

Cc'ing Cédric as co-maintainer with Alex.

Quoted 67 lines

Polite ping. Please let me know if there are any concerns with this version of the patches.

Thanks

Farhan

Quoted 59 lines · Farhan Ali

Hi,

Recently on s390x we have enabled mmap support for vfio-pci devices [1]. This allows us to take advantage and use userspace drivers on s390x. However, on s390x we have special instructions for MMIO access. Starting with z15 (and newer platforms) we have new PCI Memory I/O (MIO) instructions which operate on virtually mapped PCI memory spaces, and can be used from userspace. On older platforms we would fallback to using existing system calls for MMIO access.

This patch series introduces support the PCI MIO instructions, and enables s390x support for the userspace NVMe driver on s390x. I would appreciate any review/feedback on the patches.

Thanks
Farhan

[1] https://lore.kernel.org/linux-s390/20250226-vfio_pci_mmap-v7-0- [email protected]/

ChangeLog
---------
v4 series https://lore.kernel.org/qemu-devel/20250414213616.2675-1- [email protected]/ v4 -> v5
     - Fixup typo in PCI MMIO API (patch 2).

v3 series https://lore.kernel.org/qemu-devel/20250401172246.2688-1- [email protected]/ v3 -> v4
     - Use generic ld/st functions for non s390x PCI access suggested by Alex (patch 2).
     - Removed R-b for patch 2 as the host PCI MMIO access API changed for non-s390x. Would appreciate review on this again.

v2 series https://mail.gnu.org/archive/html/qemu-devel/2025-03/ msg06847.html v2 -> v3
     - Update the PCI MMIO APIs to reflect that its PCI MMIO access on host as suggested by Stefan(patch 2)
     - Move s390x ifdef check to s390x_pci_mmio.h as suggested by Philippe (patch 1)
     - Add R-bs for the respective patches.

v1 series https://mail.gnu.org/archive/html/qemu-devel/2025-03/ msg06596.html v1 -> v2
     - Add 8 and 16 bit reads/writes for completeness (patch 1)
     - Introduce new QEMU PCI MMIO read/write API as suggested by Stefan (patch 2)
     - Update NVMe userspace driver to use QEMU PCI MMIO functions (patch 3)

Farhan Ali (3):
   util: Add functions for s390x mmio read/write
   include: Add a header to define host PCI MMIO functions block/nvme: Use host PCI MMIO API

  block/nvme.c                  |  41 +++++----- include/qemu/host-pci-mmio.h  | 141 ++++++++++++++++++++++++++++++++ include/qemu/s390x_pci_mmio.h |  24 ++++++ util/meson.build              |   2 + util/s390x_pci_mmio.c         | 148 ++++++++++++++++++++++++++++++++++

  5 files changed, 338 insertions(+), 18 deletions(-)

  create mode 100644 include/qemu/host-pci-mmio.h create mode 100644 include/qemu/s390x_pci_mmio.h create mode 100644 util/s390x_pci_mmio.c